Appliyng filters on Haloplex produced reads

yvan

Hello, We are using some custom made and predefined Haloplex kits. I was wandering how the best practice for variant detection should be "adapted". One of the biggest challenges we are facing is that we can not do a VQSR due to the low number of variants detected. So we have to use an hard filtering step, but here again the nature of the reads, all produced by enzyme restrictions, make some filters inappropriate like the ReadPosRankSumTest as the reads are not randomly produced. I was wondering if the community has any experience with this kind of data and how the hard filtering should be made? Thanks for your help. yvan

sulinwu

Another thing to pay attention to is the ceiling of 250 calls for UnifiedGenotyper. In order to force the VQSR, there will be a sacrifice about Gaussian number and minNumber of variants. I've compared the raw and recalibrated snp calls in terms of control bams includd in the run. I would recommend to go with raw VCF at this moment before a new model tolerating/ accepting the amplicon and RE digestion nature of HaloPlex is out for variant calling.